I have binary file which has approx 77MB:

nupic@nupic-virtualbox:~/VboxSharedFolder/experiments/sync/exp2$ ls -lah src/
total 77M
drwxrwx--- 1 root vboxsf    0 Jun 21 13:31 .
drwxrwx--- 1 root vboxsf 4.0K Jun 21 16:21 ..
-rwxrwx--- 1 root vboxsf  77M May 27  2014 binary.bin

I've been playing with rsync and it's delta algorithm feature to see how it is working. Idea was to make small differences in binary file and see how much data were transferred using several methods. For those purposes I've made very simple script:

# rsync does not trnansfers delta over local by default
sed 's%\x00\x00\x00\x20\x66\x74\x79\x70\x69\x73\x6f\x6d\x00\x00\x02\x00%\x11\x11\x11\x11\x11\x11\x11\x11\x11\x11\x11\x11\x11\x11\x11\x11%' src/binary.bin > dst/binary.bin
strace -f -e trace=read,write -o rw_rsync_local_default.log rsync -avcz --progress src/ dst/

# rsync -no-W should enables delta tranfer no matter if local or remote
sed 's%\x00\x00\x00\x20\x66\x74\x79\x70\x69\x73\x6f\x6d\x00\x00\x02\x00%\x11\x11\x11\x11\x11\x11\x11\x11\x11\x11\x11\x11\x11\x11\x11\x11%' src/binary.bin > dst/binary.bin
strace -f -e trace=read,write -o rw_rsync_local_delta_enabled.log rsync --no-W -avcz --progress src/ dst/

# rsync trnansfers delta over network by default
sed 's%\x00\x00\x00\x20\x66\x74\x79\x70\x69\x73\x6f\x6d\x00\x00\x02\x00%\x11\x11\x11\x11\x11\x11\x11\x11\x11\x11\x11\x11\x11\x11\x11\x11%' src/binary.bin > dst/binary.bin
strace -f -e trace=read,write -o rw_rsync_remote.log rsync -avcz -e "ssh -o StrictHostKeyChecking=no -o UserKnownHostsFile=/dev/null" --progress src/ nupic@localhost:/home/nupic/VboxSharedFolder/experiments/sync/exp2/dst/

# scp should transfers whole file not delta
sed 's%\x00\x00\x00\x20\x66\x74\x79\x70\x69\x73\x6f\x6d\x00\x00\x02\x00%\x11\x11\x11\x11\x11\x11\x11\x11\x11\x11\x11\x11\x11\x11\x11\x11%' src/binary.bin > dst/binary.bin
strace -f -e trace=read,write -o rw_scp.log scp src/binary.bin nupic@localhost:/home/nupic/VboxSharedFolder/experiments/sync/exp2/dst/

# cp always transfers whole file not delta
sed 's%\x00\x00\x00\x20\x66\x74\x79\x70\x69\x73\x6f\x6d\x00\x00\x02\x00%\x11\x11\x11\x11\x11\x11\x11\x11\x11\x11\x11\x11\x11\x11\x11\x11%' src/binary.bin > dst/binary.bin
strace -f -e trace=read,write -o rw_cp.log cp src/binary.bin dst/binary.bin

Then I have following loop for evaluation of results:

for i in *.log; do
  echo $i; cat $i | grep write | awk 'BEGIN {FS="="}{ sum += $2} END {print sum/1024/1024 "MB"}';
  echo "###########";

Here are results:


From those five experiments only two are clear to me:

  1. cp writes approx same amount of bytes as the size of original file (rw_cp.log).
  2. rsync uses delta algorithm when destination is remote (over the network) (rw_rsync_remote.log)

And here are unclear things to me:

  1. Why invoking rsync on both src and dst on localhost writes approx two times bytes as the size of original file? (rw_rsync_local_default.log)
  2. Why --no-W option for rsync does not transfer only delta for src and dst on localhost as stated here and why it still transfers approx whole file? (rw_rsync_local_delta_enabled.log)
  3. Bonus: Why scp transfers approx twice bytes as original file size? I understand that there is some encryption but two times seems large to me (rw_scp.log).

2 Answers 2


To answer the main question in short, rsync seems to write double the number of bytes, because it spawns two processes/threads to do the copy, and there's one stream data between the processes, and another from the receiving process to the target file.

We can tell this by looking at the strace output in more detail, the process IDs in the beginning of the file, and also the file descriptor numbers in the write calls can be used to tell different write "streams" from each other.

Presumably, this is so that a local transfer can work just like a remote transfer, only the source and destination are on the same system.

Using something like strace -e trace=process,socketpair,open,read,write would show some threads spawned off, the socket pair being created between them, and different threads opening the input and output files.

A test run similar to yours:

$ rm test2
$ strace -f -e trace=process,socketpair,open,close,dup,dup2,read,write -o rsync.log rsync -avcz --progress test1 test2
$ ls -l test1 test2
-rw-r--r-- 1 itvirta itvirta 81920004 Jun 21 20:20 test1
-rw-r--r-- 1 itvirta itvirta 81920004 Jun 21 20:20 test2

Let's take a count of bytes written for each thread separately:

$ for x in 15007 15008 15009  ; do echo -en "$x: " ; grep -E "$x (<... )?write"  rsync.log | awk 'BEGIN {FS=" = "} {sum += $2} END {print sum}'  ; done 
15007: 81967265
15008: 49
15009: 81920056

Which matches pretty much with the theory above. I didn't check what the other 40kB written by the first thread is, but I'll assume it prints the progress output, and whatever metadata about the synced file rsync needs to transfer to the other end.

I didn't check, but I'll suggest that even with delta compression enabled, perhaps the "remote" end of rsync still writes out (most of) the file in full, resulting in approximately the same amount of writes as with cp. The transfer between the rsync threads is smaller, but the final output is still the same.

  • 1
    And similarly scp creates a child ssh process to do the actual SSH protocol (handshake, encryption and MAC). Thus parent reads the file and writes over pipe to child, and also writes the 'progress' meter to the terminal, which is fairly small; child reads the pipe and reads and writes the socket for the SSH protocol, which adds usually 1-2k startup and ,maybe 1% to the data. Jun 21, 2016 at 22:41
  • @ilkkachu I've tried to get written bytes per child but maybe something is wrong with regexp. Tried code: for x in $(awk '{print $1}' rsync.log | sort | uniq); do; echo -en "$x: "; grep -E "$x (<... )?write" rsync.log | awk 'BEGIN {FS=" = "} {sum += $2} END {print sum}'; done Jun 22, 2016 at 12:09

By default, rsync first creates a new copy of the target file and then replaces it, for various safety reasons. You can override this by specifying --inplace along with --no-whole-file. This tells rsync to do an in-place-edit of the target file, accepting the various risks (typically minor for this situation) as documented in the man page.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.